TF-RD-013 Shape-Aware Support Bundle

Use this support bundle when you need the committed assumptions and comparison surfaces behind TF-RD-013 issue #127.

It is a historical reference-only support surface for that shape-aware follow-up, not the main roadmap or sweep execution guide. Reconstructing these local artifacts is no longer a supported workflow.

Environment assumptions:

• TAB_FOUNDRY_ROOT is this repo root.
• DAGZOO_ROOT defaults to the sibling checkout ../dagzoo.
• The broader dagzoo follow-up uses three config-backed invocations:
  • ../dagzoo/configs/benchmark_cpu.yaml with 2048 datasets
  • ../dagzoo/configs/default.yaml with 4096 datasets
  • ../dagzoo/configs/benchmark_cuda_h100_large_shape.yaml with 128 datasets
• The curated comparator baseline remains pinned to data/manifests/bench/openml_classification_large_v1/manifest.parquet.

Historical support surface:

• materializes three explicit dagzoo generate runs under outputs/staged_ladder_support/tf_rd_013_shape_aware_dagzoo_v1/
• keeps each invocation’s handoff manifest and identity separate for provenance review
• assembles one merged dagzoo manifest with build_manifest(data_roots=[...]) and no single-handoff manifest metadata
• exports one OpenML-only curated comparator corpus under the same local support root
• writes tracked JSON summaries for the shape-aware dagzoo surface and the curated comparator surface

Committed files:

• materialization_summary.json: shape-program notes, per-invocation handoff summaries, merged-manifest assembly details, and curated comparator provenance.
• manifest_characteristics_summary.json: anchor vs shape-aware dagzoo, anchor vs curated, and shape-aware dagzoo vs curated manifest comparisons.

Local-only files:

• Generated dagzoo shards, curated OpenML packed shards, and local manifests lived under outputs/staged_ladder_support/tf_rd_013_shape_aware_dagzoo_v1/.
• outputs/ stays ignored because the dagzoo artifacts are too large to commit as repo-tracked fixtures.

Policy notes:

• This bundle is the issue #127 follow-up to the neutral first promoted-anchor TF-RD-013 read from issue #122.
• The broader dagzoo read is still about synthetic-data coverage, not filtering policy; issue #124 remains a later question only if the shape-aware result exposes a specific predictability problem.
• The curated comparator remains OpenML-first and evidence-only unless later approved augmentations are explicitly justified.